Methodological report for "Do SMEs

create more and better jobs?"

Jan de Kok Wim Verhoeven Niek Timmermans Ton Kwaak

Zoetermeer, November 2011

EIM Business & Policy Research P.O. Box 7001 2701 AA Zoetermeer, The Netherlands Phone: +31 79 3430200 Fax: +31 79 3430203 Email: pvr@eim.nl Website: www.eim.nl EIM Office in Brussels 5, Rue Archimède, Box 4, 1000 Brussels Phone: +32 2 5100884 Fax: +32 2 5100885 Email: rvh@eim.nl

The views expressed herein are those of the consultant, and do not represent the official views of the Commission. This report was developed by EIM Business & Policy Research (the Netherlands) in the context of the framework contract ENTR/2007/040-1 for the provision of Economic Studies in support of SME policy. The project was managed by Paul Vroonhof of EIM. The core project team consisted of Jan de Kok, Wim Verhoeven, Niek Timmermans, Ton Kwaak, Jacqueline Snijders and Florieke Westhof. EIM bv is entriely responsible for the contents of this report. Quoting numbers or text in papers, essays and books is permitted only when the source is clearly cited. No part of this publication may be copied and/or published in any form or by any means whatsoever, or stored in a retrieval system, without the prior written permission of EIM bv. EIM bv accepts no responsibility for printing errors and/or other imperfections.

This report was prepared by EIM Business & Policy Research with financial support from the European Communities, under the Competitiveness and Innovation Programme 2007-2013.

3

Contents

1 Introduction 5

2 Orbis-Amadeus 7 2.1 Characteristics of the data 7 2.2 Preparation of the complete database 8 2.3 Preparation of the restricted database 14

3 Structural Business Statistics SBS 17 3.1 Characteristics of the data 17 3.2 Measuring employment growth by enterprise size class 17 3.3 Discussion of other methodologies applied 20

4 Enterprise Survey 2010 21 4.1 Questionnaire 21 4.2 Stratification plan 24 4.3 Survey 33 4.4 Preparation of the dataset 36 4.5 Analysis 42

Annex

I Enterprise Survey 2010 Questionnaire 45

5

1 Introduction

On 3 March 2010, the Commission launched the "Europe 2020 Strategy for smart, sustainable and inclusive growth". Europe 2020 is the EU's growth strat-egy for the coming decade, which entails transforming itself into a smart, sus-tainable and inclusive economy leading to high levels of employment, productiv-ity and social cohesion. The strategy presents concrete actions to be taken at the EU and the national levels. Smart growth refers to fostering knowledge, innova-tion, education and digital society. Sustainable growth refers to making EU pro-duction more resource efficient while improving competitiveness, and inclusive growth focuses on raising participation in the labour market, the acquisition of skills and fighting poverty. The Small Business Act (SBA) for Europe was adopted in June 2008. The SBA es-tablishes a comprehensive SME policy framework for the EU and its Member States. The SBA review was presented in February 2011. The review includes the progress of the implementation of the SBA and new actions to be taken by the EC and Member States to respond to challenges resulting from the economic cri-sis. Building on work carried out by the Commission and external actors, the Com-mission has launched a study to investigate the role SMEs play in job creation and the quality of jobs they provide, particularly in the light of the SBA Act and the Europe 2020 strategy. The results of this study are presented in the publica-tion "Do SMEs create more and better jobs?" The results in the study are based primarily on three different data sources: the Orbis-Amadeus database, the Structural Business Statistics (SBS) and the Enterprise Survey 2010. These data sources and the methodologies applied to them are discussed in this document. The following three chapters discuss the details of each of these data sources and the methodologies applied. In the remainder of the introduction, some gen-eral observations are made concerning the comparability of the results of these sources.

Different sector and size c lass c lassi f icat ions used

The publication "Do SMEs create more and better jobs?" focuses on the business economy, which is defined by NACE Sections1 D, F-K, N and O (excl. 91). Many of the figures and tables presented in the first chapters of "Do SMEs create more and better jobs?" are, however, based on the publication "European SMEs under Pressure", which uses a different sectoral demarcation: that of the non-financial business economy, defined by NACE Sections C -I and K. Notice that the differ-ence between the non-financial business economy and the business economy is not limited to financial intermediation (NACE J), but also differs regarding private enterprises from other service activities (NACE N health and social work and NACE O other service activities, excl. 91)2.

1 This refers to NACE rev. 1.1.

2 In addition, they also differ regarding mining and quarrying (NACE C) and electricity, gas and water supply (NACE E), but these two divisions involve relatively few private enterprises.

6

Furthermore, the publication "Do SMEs create more and better jobs?" is con-cerned with jobs of employees. Although technically speaking, entrepreneurs and the self-employed also have jobs, the common understanding of a job is limited to jobs of employees. Hence, enterprises without employees were excluded from the Enterprise Survey 2010, and all the results based on the Enterprise Survey 2010 refer to employer enterprises only (defined as enterprises employing at least one employee). Unfortunately, many other available databases with enter-prise statistics do not distinguish between employer enterprises (enterprises em-ploying at least one employee) and enterprises without employees (self-employed entrepreneurs). Most of the information presented in Part A of "Do SMEs create more and better jobs?" refers to all jobs, i.e. including the employ-ment of self-employed and non-paid family workers. The information presented in Part B usually refers to jobs of employees only.

Data availabil ity for countr ies outside the EU varies between data sources

The study covers the 27 Member States of the European Union and the following 10 non-EU countries: Albania, Croatia, the Former Yugoslav Republic of Mace-donia, Iceland, Israel, Liechtenstein, Montenegro, Norway, Serbia and Turkey. However, not all the available data sources cover all countries. The Structural Business Statistics only cover the EU27 Member States1. The Orbis-Amadeus da-tabase includes statistics on 24 of the 27 Member States (Cyprus, Malta and Luxembourg are excluded because too few observations are available), 5 of the 10 non-EU countries (Croatia, Iceland, Liechtenstein, Norway and Serbia) and Switzerland. The Enterprise Survey 2010 includes all 37 countries.

1 The SBS also includes data on Norway, but not on any of the other non-EU countries that partici-

pated in this study. The SBS data for Norway is therefore not used for this study.

7

2 Orbis-Amadeus

2.1 Characteristics of the data

Orbis and Amadeus

The Orbis-Amadeus database is based on two databases, Orbis and Amadeus, which are built and maintained by Bureau Van Dijk (an organisation that special-ises in providing company information). The Orbis database contains general in-formation on more than 40 million enterprises in Europe. It includes the follow-ing information1, amongst others: − general contact information; − summary of company details, including number of employees, age of com-

pany, legal form, industry and activities (including primary and secondary codes in several local and international classifications);

− financial statistics, in specific formats for corporate, banking and insurance companies;

− ownership information. The Amadeus database contains even more detailed information for 14 million European enterprises. For this study a subset of all enterprises from the Orbis and Amadeus databases has been used. This subset includes enterprises that: − are included in either the Amadeus or the Orbis database (or in both); − are located in the EU, in Switzerland or in one of the non-EU countries that

are covered by this study (see Chapter 1); − are active in the business economy; − existed at the end of 2008; − employed fewer than 250 employees at the end of 2008; − and for which employment statistics were available for 2008. The large majority of the enterprises included in the Orbis and Amadeus data-bases do not meet these criteria (in particular the criteria that the enterprise had to exist at the end of 2008, that employment information for 2008 had to be available, and that the enterprises should be located in one of the 37 European countries of this study). Nevertheless, the resulting Orbis-Amadeus database contains observations on somewhat fewer than 3 million enterprises.

Micro enterprises under-represented in Orbis-Amadeus

To a large extent, the Orbis-Amadeus database is based on register data that is obtained from local Chambers of Commerce and/or tax agencies. Micro enter-prises in particular are not always obliged to provide detailed company informa-tion (including employment levels). As a result, the Orbis-Amadeus database contains information on most of the small and medium-sized enterprises, but on a smaller share of all micro enterprises. This is especially true for enterprises with no employees or only one employee. These enterprises account for a signifi-cant part of the actual SME population, but the Orbis-Amadeus database contains

1 More information on this database can be found on the website of Bureau Van Dijk,

www.bvdinfo.com.

8

relatively few enterprises with fewer than 2 employees (the Orbis-Amadeus da-tabase includes more than 600,000 enterprises with fewer than 2 employees, which is in all respects a very large sample. Nevertheless, compared to the lar-ger enterprises in this database, this is a relatively small share of the total en-terprise population). In the analyses it is assumed that the developments of en-terprises with 2-9 persons employed are representative for all micro enterprises.

Usage of Orbis-Amadeus

The complete Orbis-Amadeus database can be used to present a picture of the employment situation for 2008. The preparations of this complete database are discussed in Section 2.2. For many enterprises, the Orbis-Amadeus database also includes data on the employment history. This offers the possibility of de-termining patterns of employment changes over time, at the level of individual enterprises. The additional preparations of this restricted database are discussed in Section 2.3. It should be kept in mind that any patterns revealed by this re-stricted version of the Orbis-Amadeus database refer to a specific subset of en-terprises, namely enterprises that still existed at the end of 2008. An analysis of employment changes during 2004-2008 therefore includes enterprises that started during this period (as long as they survived until the end of 2008), but excludes enterprises that did not survive during this period.

2.2 Preparation of the complete database

The enterprises in the Orbis-Amadeus database have been classified into differ-ent categories, based on country, sector and age.

2.2.1 Classi f icat ion by country groups

Classi f icat ion by EU Membership status

Enterprises are located in individual countries, which in turn can be classified into different country groups. The standard country classification that is used for this study is based on the country's relationship with the European Union. Coun-tries are classified into one of the following three country groups: 1 EU15 (the 15 original Member States of the EU). 2 EU12 (the 12 Member States that joined the EU after 1990). 3 Non-EU countries (these include the following 6 countries: Switzerland, Croa-

tia, Iceland, Liechtenstein, Norway and Serbia. Date was not available for Montenegro, Albania, Israel, the Former Yugoslav Republic of Macedonia and Turkey).

This classification can be reduced to a classification of two groups: 1 EU27 (the 27 Member States of the European Union on 2010); 2 Non-EU countries. There are many other criteria that can be used to classify countries into different groups. For this study, countries have also been classified based on their relative competitiveness, innovation performance and size-class dominance. The result-ing country groupings did not provide much additional insight. The results of these alternative classifications are therefore not included in the main report.

9

Classi f icat ion by compet it iveness

The grouping of countries by competitiveness is based on the Global Competi-tiveness Report 2010-2011 from the World Economic Forum. The (unweighted) average score on the Global Competitiveness Index for EU-27 is used as a threshold (4.7). Countries for which the Global Competitiveness Index is above this threshold are classified as highly competitive countries, and the remaining countries are classified as less competitive countries (Table 1).

Table 1 Classification of countries by competitiveness

Highly competitive countries Less competitive countries

Country Competitive Country Competitive

Switzerland 5.63 Iceland 4.68

Sweden 5.56 Estonia 4.61

Germany 5.39 Czech Republic 4.57

Finland 5.37 Poland 4.51

Netherlands 5.33 Cyprus 4.5

Denmark 5.32 Spain 4.49

United Kingdom 5.25 Slovenia 4.42

Norway 5.14 Lithuania 4.38

France 5.13 Portugal 4.38

Austria 5.09 Italy 4.37

Belgium 5.07 Montenegro 4.36

Luxembourg 5.05 Malta 4.34

Ireland 4.74 Hungary 4.33

Slovakia 4.25

Romania 4.16

Romania 4.16

Latvia 4.14

Bulgaria 4.13

Croatia 4.04

the Former Yugoslav Republic

of Macedonia 4.02

Greece 3.99

Albania 3.94

Serbia 3.84

Source: Global competitiveness report 2010-2011, World Economic Forum.

Classi f icat ion by size-class dominance

A country is said to be micro, SME, or LSE dominated if either micro, small and medium-sized (taken together), or large-scale enterprises have the largest share in total employment of the non-financial business economy. The majority of countries are SME dominated; six countries are micro-dominated and five coun-

10

tries are LSE-dominated (the size-class dominance could not be determined for Turkey).

Table 2 Classification of countries by size class dominance

Size class dominance

micro small and medium-sized LSE

Greece Austria Finland

Italy Belgium France

Portugal Denmark United Kingdom

Malta Germany Slovakia

Poland Ireland Iceland

Albania Luxembourg

Netherlands

Spain

Sweden

Bulgaria

Cyprus

Czech Republic

Estonia

Hungary

Latvia

Lithuania

Romania

Slovenia

Norway

Switzerland

Liechtenstein

Croatia

the Former Yugoslav Republic of Macedonia

Montenegro

Serbia

Note: The size-class dominance could not be determined for Turkey.

Source: Own calculations, based on Structural Business Survey.

Classi f icat ion by innovation performance

Information on the innovation performance of countries is based on the Innova-tion Union Scoreboard (IUS) 2010, published by DG Enterprise and Industry. The IUS collects information on 25 different indicators of innovation. These indicators are classified into three main types and eight dimensions: − Type one: enablers

− Human resources (3 indicators) − Open, excellent and attractive research systems (3 indicators) − Finance and support (2 indicators)

11

− Type two: firm activities − Firm investments (2 indicators) − Linkages and entrepreneurship (3 indicators) − Intellectual assets (4 indicators)

− Type three: outputs − Innovators (3 indicators) − Economic effects (5 indicators)

The data used for the IUS 2010 relate to actual performance in 2007 (4 indica-tors), 2008 (10 indicators) and 2009 (10 indicators). As a consequence, the IUS 2010 does not fully capture the impact of the financial crisis on innovation per-formance. The scores on the different indicators are used to construct a sum-mary innovation index (SSI). Based on the scores on this index, countries are classified into four different categories of innovation performance (Table 3).

Table 3 Innovation performance of individual countries

Modest innovators Moderate innovators Innovation followers Innovation leaders

Bulgaria Croatia Austria Denmark

Latvia Czech Republic Belgium Finland

Lithuania Greece Cyprus Germany

Romania Hungary Estonia Sweden

Serbia Italy France

Turkey Malta Iceland

the Former Yugoslav

Republic of Macedonia

Poland Ireland

Portugal Luxembourg

Slovakia Netherlands

Spain Norway

Slovenia

United kingdom

Note: The innovative performance is not determined for Albania, Montenegro, Israel and

Liechtenstein.

Source: Innovation Union Scoreboard (IUS) 2010, Annex G.

2.2.2 Classi f icat ion by sector

Seven broad areas of economic activity are distinguished1: 1 Manufacturing (NACE D). 2 Construction (NACE F). 3 Wholesale trade (NACE 51). 4 Retail trade (NACE 50, 52). 5 Transport and communication (NACE I). 6 Business services (NACE J and K). 7 Personal services (NACE H, N and O).

1 The sample size did not allow distinguishing more sectors.

12

Together, these seven sectors define the business economy (NACE D, F-K, N and O). The division NACE 91 (Activities of membership organisations, such as profes-sional organisations, trade unions and political organisations) are included as part of the personal services, and hence as part of the business economy. In the Structural Business Statistics (SBS) and the Enterprise Survey 2010, NACE Divi-sion 91 is excluded from the definition of personal services (and, hence, of the business economy). Business demography indicators from the SBS indicate that within the EU, NACE Division 91 includes only a relatively small number of enter-prises1, so whether this section is included or not will only have a very small ef-fect on aggregate findings.

2.2.3 Classi f icat ion by age

The age of an enterprise can be classified into three groups: 1 Newly born enterprises (up to five years old). 2 Young enterprises (5 up to 10 years old). 3 Established enterprises (10 years and older). The resulting complete Orbis-Amadeus database includes 2.9 million SMEs from 30 European countries. For the 27 Member States of the EU, 2.6 million observa-tions are available. This database is used to prepare the tables regarding the employment situation in 2008. The number of observations varies by industry, but especially by country. For Cyprus, Malta and Luxembourg the number of ob-servations is actually so low that these countries are not included in the data-base (see Table 4). The Orbis-Amadeus database is therefore mainly used for analysis and tables on a high aggregate level (such as comparing EU12, EU15 and non-EU). Results pertaining to individual countries should be interpreted with care.

1 The enterprises in NACE Division 91 account for 6% of all enterprises in NACE Section O and 2%

of all enterprises in the personal services.

13

Table 4 Available number of enterprises with employment data in 2008, by country and

sector

Country manufacturing construction

wholesale

trade retail trade

transport and

communication

business

services

personal

services total

Austria 8,199 8,448 6,393 12,360 3,637 26,711 8,355 74,103

Belgium 14,038 18,768 15,760 21,485 6,419 26,332 19,789 122,591

Bulgaria 1,198 552 1,540 981 555 712 339 5,877

Czech Re-

public 4,371 2,166 4,999 3,894 871 6,522 1,594 24,417

Germany 24,000 15,234 20,211 37,220 9,939 51,371 15,046 173,021

Denmark 6,385 7,690 7,431 6,874 2,369 13,106 4,567 48,422

Estonia 3,151 3,412 2,839 3,012 2,238 4,785 2,110 21,547

Spain 45,003 45,101 34,309 39,155 13,321 59,823 28,627 265,339

Finland 5,051 4,730 3,092 3,604 2,552 7,260 3,268 29,557

France 70,108 90,410 46,798 103,532 21,335 102,474 92,189 526,846

United King-

dom 8,941 3,946 5,207 3,150 2,536 19,064 10,074 52,918

Greece 4,539 935 4,928 1,870 819 1,845 1,951 16,887

Hungary 2,221 971 1,839 1,518 1,142 1,136 277 9,104

Ireland 2,363 1,465 1,670 1,830 482 2,098 582 10,490

Italy 80,741 44,559 40,250 36,132 14,984 58,144 33,518 308,328

Lithuania 6,924 4,745 5,942 14,077 5,783 13,034 8,898 59,403

Latvia 5,550 4,797 5,264 11,532 4,157 13,851 6,277 51,428

Netherlands 9,762 9,012 13,644 7,657 4,125 23,637 4,457 72,294

Poland 2,805 822 2,214 875 393 969 260 8,338

Portugal 31,299 27,420 24,280 43,741 15,157 38,186 37,883 217,966

Romania 40,602 35,105 34,651 84,664 25,687 66,557 31,075 318,341

Sweden 21,293 23,131 17,382 23,304 11,544 58,521 20,431 175,606

Slovenia 2,482 735 2,274 904 690 741 157 7,983

Slovakia 857 445 832 572 191 782 353 4,032

Total EU 27 401,883 354,599 303,749 463,943 150,926 597,661 332,077 2,604,838

Switzerland 22,763 22,681 22,410 29,949 6,146 61,813 16,624 182,386

Croatia 5,116 3,555 5,497 4,372 1,692 4,748 2,366 27,346

Iceland 286 536 196 296 140 549 382 2,385

Liechtenstein 157 67 84 125 31 452 73 989

Norway 8,851 11,978 8,632 15,394 4,613 25,506 12,229 87,203

Serbia 3,863 1,158 3,936 1,387 863 1,291 585 13,083

Total non-EU

countries 41,036 39,975 40,755 51,523 13,485 94,359 32,259 313,392

Total Orbis-

Amadeus

database 442,919 394,574 344,504 515,466 164,411 692,020 364,336 2,918,230

14

Source: Complete Orbis-Amadeus database, based on data from Bureau van Dijk.

Results weighted to correct for under-representation of smaller enterprises

Micro enterprises in particular are under-represented in the Orbis-Amadeus da-tabase. Therefore, the actual number of jobs in SME will be higher than the number of SME jobs determined from the Orbis-Amadeus database. All tables that present employment statistics based on the Orbis-Amadeus database there-fore present weighted results, where the weights are based on statistics from the Structural Business Statistics regarding the total numbers of SMEs by size class and sector1. This ensures that the overall employment dynamics are representa-tive for the SME population.

2.3 Preparation of the restricted database The Orbis-Amadeus database is mainly used to determine patterns of employ-ment changes over time. This requires a database that contains annual employ-ment information on enterprises for a certain period. The restricted database re-fers to those enterprises from the complete database, for which the required an-nual employment information is available.

2.3.1 Period under investigation

In theory, the Orbis and Amadeus databases should contain a track record of the enterprises over 10 years. In practice, the track record of enterprises in the Or-bis-Amadeus database only holds for the period of the last 4 to 5 years. A com-plete track record over the past 10 years is only available for a few percent of the enterprises older than 10 years. This makes it very difficult to perform an unbiased analysis of the employment changes of established SMEs for a period of 10 years. The analysis is therefore limited to an analysis of employment changes over five years, in particular the period 2004-2008.

2.3.2 Deal ing with missing data

For some enterprises, information on the number of persons employed is not available for every year of the analysis period. If employment data on the enter-prise level is missing for 2005, 2006 and/or 2007, interpolations are used to provide estimates of the missing data. In the case where employment data is missing for 2004, the enterprise is removed from the sample2. This is the case for a substantial number of the enterprises from the complete database. Once these are removed, the restricted database includes employment information on 2004 and 2008 for just over one million enterprises. The majority of these enter-prises were established during the period under investigation (from 1-1-2004 onwards); about a quarter of the enterprises in the database (994,000) were es-tablished before 1-1-2004 (and still existed on 31-12-2008).

1 The weights are calculated by dividing the actual level of SME jobs according to the Structural

Business Statistics (within each sector and size class) by the number of SME jobs in the Orbis-Amadeus database. The weights are computed for 2008 and are used back to 2004.

2 Employment information on 2008 is always available; otherwise, the enterprise would not be included in the complete database.

15

2.3.3 Classi f icat ion by employment growth type

The enterprises in the restricted database are classified into three different growth types: growing, stable and shrinking enterprises. The classification is based on a comparison of the employment levels at the beginning and the end of the period; developments during the period are ignored. To be precise, the fol-lowing classification rules are applied to all young and established enterprises (all enterprises aged 5 years or more): 1 Growing enterprises: enterprises where the number of employees at the end

of 2008 is higher than the number of employees at the end of 2004. 2 Stable enterprises: enterprises where the number of employees at the end of

2008 is exactly equal to the number of employees at the end of 2004. 3 Shrinking enterprises: enterprises where the number of employees at the end

of 2008 is lower than the number of employees at the end of 2004. To remove outliers, maximum growth and shrink rates have been determined for each size class. Enterprises that fall outside the following boundaries are re-moved from the database: Size class 1 (1-9 employees in 2004) − A maximum growth rate of 4906% for the period 2004-2008 (based on an an-

nual growth rate of 166% for each year). This limit allows an employment growth from 1 to 50 or from 9 to 451.

− No maximum employment decline is used. A decline from 9 employees to 1 is allowed.

Size class 2 (10-49 employees in 2004) − A maximum growth rate of 1500% for the period 2004-2008 (based on an an-

nual growth rate of 100% for each year). This limit allows an employment growth from 10 to 160 or from 49 to 784.

− No maximum employment decline is used. A decline from 49 employees to 1 is allowed.

Size class 3 (50-249 employees in 2004) − A maximum growth rate of 1500% for the period 2004-2008 (based on an an-

nual growth rate of 100% for each year). This limit allows an employment growth from 50 to 800 and from 249 to 3984.

− A maximum employment decline of -94% for the period 2004-2008 (based on an annual decline of 50% for each year). A decline from 249 to 16 or from 50 to 3 is therefore possible.

The database also contains apparent mistakes, probably due to erroneous repro-duction from other databases or for other reasons. These data have also been removed. The restricted Orbis-Amadeus database includes 1.0 million SMEs from 24 EU countries. No data are available for Cyprus, Malta and Luxembourg. For the non-EU countries, the number of observations in the restricted Orbis-Amadeus data-base is so small that it is not possible to present average statistics (not even at the level of the non-EU countries as a group). The restricted Orbis-Amadeus da-tabase therefore only includes enterprises from the EU. This database is used for the analyses regarding employment changes of individual firms from 2004 to 2008 (Table 5).

16

Table 5 Available number of enterprises with employment data for 2004 and 2008, by

country and sector

Country* manufacturing construction

wholesale

trade

retail

trade

transport and

communication

business

services

personal

services total

Austria 3,479 3,466 3,068 4,067 1,391 2,573 1,074 19,118

Belgium 10,844 12,520 11,457 14,187 4,667 14,184 9,162 77,021

Bulgaria 940 374 1,077 619 313 446 243 4,012

Czech Republic 2,535 1,117 2,490 2,019 433 2,807 693 12,094

Germany 7,814 4,530 6,579 14,564 3,031 7,904 2,657 47,079

Denmark 3,677 3,499 4,259 3,381 1,218 5,730 2,166 23,930

Estonia 2,302 1,644 1,880 2,238 1,432 2,649 1,398 13,543

Spain 32,973 26,732 23,766 25,745 8,518 31,124 15,362 164,220

Finland 3,509 2,692 2,013 2,263 1,613 3,751 1,827 17,668

France 37,140 39,037 23,425 46,809 10,218 40,562 36,149 233,340

United Kingdom 6,232 2,370 3,646 2,171 1,558 10,056 6,119 32,152

Greece 3,896 703 3,924 1,490 645 1,486 1,670 13,814

Hungary 56 12 30 20 17 7 2 144

Ireland 4 0 0 1 0 3 0 8

Italy 22,190 5,618 8,513 5,730 2,878 8,522 4,083 57,534

Lithuania 339 215 526 209 110 73 14 1,486

Latvia 2,765 1,213 2,155 5,919 1,386 4,870 2,958 21,266

Netherlands 808 468 932 237 253 631 117 3,446

Poland 2,140 593 1,517 615 292 671 161 5,989

Portugal 407 156 289 172 63 178 98 1,363

Romania 25,552 11,893 19,477 51,862 11,843 28,862 16,238 165,727

Sweden 17,147 16,478 13,113 17,080 8,803 37,129 13,397 123,147

Slovenia 1,931 502 1,783 661 434 481 102 5,894

Slovakia 193 74 182 62 20 75 35 641

total 188,873 135,906 136,101 202,121 61,136 204,774 115,725 1,044.636

* No data available for Cyprus, Malta and Luxembourg.

Source: Restricted Orbis-Amadeus database, based on data from Bureau van Dijk.

17

3 Structural Business Statistics SBS

3.1 Characteristics of the data

The Structural Business Statistics (SBS) database is associated with the Annual Report 2009 of the EC's SME Performance Review1. This database provides sta-tistical data on the number of enterprises and employment by country, enter-prise size class and sector of industry, amongst other data. Countries include all EU Member States. Size classes distinguished are micro, small, medium-sized and large enterprises, according to a head count of the number of persons em-ployed in the year the data relate to. Sectors of industry include all NACE Rev. 1.1 divisions within the non-financial business economy, i.e. NACE C -I and K.

3.2 Measuring employment growth by enterprise size class

The SBS data is used to determine the extent to which enterprises from different size classes have contributed to net employment changes, amongst other things. This requires that the available longitudinal data on employment levels by size class be corrected for the so-called population effect. This section describes the nature of the population effect and why it is important to correct for it.

The population effect

At the macro level as well as by sector of industry, employment growth is the balance of job creation on the one hand and job destruction on the other. Job creation and destruction may occur because of employment change in incumbent enterprises, or because of entry and exit of enterprises. An enterprise size-class is defined as a population of enterprises that falls within certain size-class boundaries at a specific point in time. As indicated above, a distinction is made between micro, small, medium-sized and large enterprises, based on a head count of persons employed. Available data measure employ-ment by size class in a certain year as the total employment in all enterprises that belong to that size class in that particular year. By comparing such employment figures for, say, SMEs, in two years t0 and t1, one compares the number of employed persons in enterprises in the SME popula-tion in t0 with the number of employed persons in enterprises in the SME popula-tion in t1. However, enterprises can cross size-class boundaries at any time. The incumbent SMEs at time t1 may have been large enterprises at time t0; con-versely, enterprises that had fewer than 250 persons employed at time t0 (and thus were SME at that time) may have become large enterprises at time t1. Comparing employment figures for SMEs in the two years t0 and t1 therefore in-cludes the impact of previously large enterprises that became SME (positively af-

1 European Commission: European SMEs under Pressure: Annual Report on EU Small and Medium-

sized Enterprises 2009 (http://ec.europa.eu/enterprise/policies/sme/facts-figures-analysis/performance-review/pdf/dgentr_annual_report2010_100511.pdf ).

18

fecting the measured employment change in the SME population), as well as the impact of enterprises that previously were SMEs that have become large (nega-tively affecting the measured employment change in the SME population). This example shows that changes in the employment level of a certain size class can be attributed to either one of two different causes: − changes in the level of employment of individual enterprises: job creation and

destruction by enterprises (including entry and exit of enterprises); − changes in the classification of enterprises in size classes: the population ef-

fect. Once the impact of one of these causes is known or estimated, the impact of the other can be calculated by subtracting the known effect from the measured em-ployment change in the size class under review.

Decomposing employment changes by size c lass: adjust ing for the po-pulat ion effect

To what extent have enterprises from different size classes contributed to gross or net employment changes? To what extent are employment changes caused by enterprises from different sizes? Because of the population effect, this question cannot be answered by looking at changes in the employment level of size classes. An adjustment must be made for changes in employment statistics that are due to enterprises crossing size-class boundaries. One way to adjust for the population effect is to classify each enterprise in a sin-gle size class for the period between two measurements. This corrects for the population effect, since the number of enterprises in each size class is now con-stant during that period. The question that remains is how individual enterprises should be classified into size classes. In the literature1 one or more of the follow-ing three classification methods have been applied: 1 Classification by initial size: the size of the enterprise at the beginning of

each measurement period determines the size class in which the enterprise is classified;

2 Classification by end size: the size of the enterprise at the end of each meas-urement period determines the size class in which the enterprise is classified;

3 Classification by average size: the average of the enterprise's size at the be-ginning and the end of each measurement period determines the size class in which the enterprise is classified.

The method of Classification by current size takes a different approach: instead of classifying enterprises, this method directly classifies individual employment changes, based on the size of the enterprise prior to each individual change. Ba-sically, it works as follows: if an enterprise employs 230 persons at t0 and 255 persons at t1, then its employment increase from 230 to 250 is attributed to SMEs, while the employment increase from 250 to 255 is attributed to large en-terprises. Classification by current size solves a series of problems associated with (some of) the other methods. In the first place, classifications by initial size and by end

1 See the literature review in De Kok, J., G. de Wit and K. Suddle (2006), SMEs as job engine of

the Dutch private economy: A size-class decomposition of employment changes for different sec-tors of the Dutch economy, EIM, 2006, Research Report H200601.

19

size share the problem that the results are affected by fluctuations around size class boundaries (see text box). This problem does not occur with classification by current size (nor with classification by average size). In the second place, if enterprises are classified into specific size classes (as is the case with the first three methods), the results depend on how often employment is measured. For example, suppose that enterprise A employs 9 employees in January, 11 em-ployees in June, and 15 employees the following January. If for instance em-ployment was only measured each January, classification by initial size would at-tribute an employment growth of +6 to the size class of micro enterprises. If employment was measured on a semi-annual basis (e.g., January and July), classification by initial size would attribute an employment growth of +2 to the size class of micro enterprises and an employment growth of +4 to the size class of small and medium-sized enterprises. Classification by current size does not suffer from this measurement bias because it does not involve classifying enter-prises into specific size classes. In the third place, classification by current size is the only classification method that can be used to decompose net employment changes by size classes without using micro data.

Impact of enterprises crossing size boundaries on the size-class pattern of employment

according to four classification methods: an example

The size of enterprise B fluctuates around the size-class boundary of 10 employees. In the first

year employment increases from 9 to 14 employees, while in the second year employment drops

back to 9 employees.

− Classification by initial size attributes an employment increase of +5 to the size class of micro

enterprises in the first year, while the employment decrease of -5 in the second year is at-

tributed to the size class of small enterprises.

− Classification by end size attributes an employment increase of +5 to the size class of small

enterprises in the first year, while the employment decrease of -5 in the second year is at-

tributed to the size class of micro enterprises.

− Classification by average size attributes the employment increase of the first year as well as

the employment decrease of the second year to the size class of small enterprises.

− Classification by current size attributes an employment increase of +1 to the size class of mi-

cro enterprises and an employment increase of +4 to the size class of small enterprises, and

for the second year attributes an employment decrease of -4 to the size class of small enter-

prises and an employment decrease of -1 to the size class of micro enterprises.

In case of both classification by initial size and by end size, the size class fluctuations of enter-

prise B do not cancel out over the two years together.

Because of these advantages, classification by current size is used here to adjust for the population effect.

Estimat ing the populat ion effect using classi f icat ion by current s ize

Statistics on changes in the number of enterprises per size class are used to cor-rect for the population effect. The appropriate adjustments are presented in Table 6. A formal proof of these adjustments can be found elsewhere (see previ-ous footnote). Here an intuitive argument is presented. For the largest size class boundary, the net number of times that the boundary between medium-sized and large enterprises is crossed is equal to the change in the number of large

20

enterprises: •Nl1. Hence, the adjustment for the size class of medium-sized

(large) enterprises is plus (minus) •Nl (the net number of crossings) times 250 (employment size at the boundary). It is plus 250⋅ •Nl for medium-sized enter-prises because the (net) change in the number of large enterprises relates to en-terprises that previously were assigned to size class of medium-sized, and minus 250⋅ •Nl because previously these enterprises were not yet large enterprises. Similarly, it is clear that the net number of times that the boundary between small and medium-sized enterprises is crossed must be equal to the change in the number of medium-sized and large enterprises together: •Nms+l. Hence, the adjustment for the size class of small (medium-sized) enterprises is •Nms+l times 50. The adjustment for the size class of micro (small) enterprises is •Ns+ms+l times 10.

Table 6 Required adjustment for the population effect (classification by current size)

Size class Adjustment

for lower boundary for upper boundary

Micro enterprises (1-9) 0 +10· •Ns+ms+l

Small enterprises (10-49) -10· •Ns+ms+l +50 · •Nms+l

Medium-sized enterprises (50-249) -50 ·•Nms+l +250·•Nl

Large enterprises (>= 250) -250·•Nl 0 (not applicable)

•NX+Y: the net change in the total number of enterprises in size classes x+y; "s" represents small

enterprises; "ms" represents medium-sized enterprises; "l" represents large enterprises.

Notice that the value of the adjustment does not depend on the initial or end size of enterprises crossing boundaries. This allows for calculating the adjustment for all enterprises together without the help of micro data.

3.3 Discussion of other methodologies applied

The database associated with the Annual Report 2009 in the framework of the European Commission's SME Performance Review covers the period 2002-2008 (and a forecast for 2009-2011). For EU15 countries, data on employment and the number of enterprises are also available from the various Observatory of European SMEs reports published by the European Commission2. These data have been linked to the data from the Annual Report database under the assumption that the Observatory of European SMEs data adequately describe trends on the number of enterprises and employment. By doing so, a series on employment and the number of enterprises by sector of industry and enterprises size class was established covering 1988-2008 for EU15 countries.

1 Entries and exits of large enterprises are treated as enterprises that move through all size

classes and pass all size class boundaries in the process. Changes in the number of large enter-prises are therefore due to crossings of the boundary between medium-sized and large enter-prises only.

2 See in particular European Commission: SMEs in Europe 2003 (http://ec.europa.eu/enterprise/policies/sme/files/analysis/doc/smes_observatory_2003_report7_en.pdf).

21

4 Enterprise Survey 2010

4.1 Questionnaire

4.1.1 Unit of observat ion

The European Commission has defined the size class of micro, small and me-dium-sized enterprises (SMEs) as all enterprises that employ fewer than 250 persons, have an annual turnover not exceeding EUR 50 million, and/or an an-nual balance sheet total not exceeding EUR 43 million1. For statistical purposes, however, a different definition of SMEs is generally used, in which SMEs are de-fined as enterprises employing fewer than 250 people. Large enterprises are de-fined as those with 250 or more persons employed2. This definition is used in various databases (including the SBS data) and publications (including Eurostat publications3 and the Observatories of European SMEs). This statistical definition is also used for the Enterprise Survey 2010. The size class of SMEs contains enterprises and not local units or enterprise groups. The observational unit for the Enterprise Survey 2010 is therefore the enterprise. In the case of enterprise groups, this implies that the survey targets the enterprises (subsidiaries) that make up the enterprise group, rather than the enterprise group as a whole.

Autonomous, partnered and l inked enterprises

Enterprises may have various relations with other enterprises. A distinction can be made between autonomous, partnered and linked enterprises. Autonomous enterprises are totally independent enterprises and enterprises holding less than 25% of the capital or voting rights (whichever is the higher) in one or more other enterprises4. This is by far the largest category. The other two categories are partnered and linked enterprises. In the case of partnered or linked enterprises, the size of the other enterprises involved (in terms of number of employees, turnover and balance sheet) should be combined with the size of the particular enterprise, before determining its size class5. In the case of a telephone survey this is not a feasible solution. In-stead, the survey includes a question that distinguishes between autonomous enterprises on the one hand, and partnered or linked enterprises on the other hand. This allows the researcher to check the extent to which the answers to various questions differ between autonomous and non-autonomous enterprises.

1 Source: Commission Recommendation of 6 May 2003 concerning the definition of micro, small

and medium-sized enterprises (2003/361/EC), Official Journal of the European Union L124/36.

2 http://epp.eurostat.ec.europa.eu/statistics_explained/index.php/Large_enterprises.

3 For example, see Statistics in Focus on SMEs and entrepreneurs in the EU http://ec.europa.eu/eurostat/ramon/coded_files/KS-NP-06-024-EN.pdf.

4 The opposite must also apply: outsiders may not have a stake of 25% or more of the capital or voting rights (whichever is the higher) in the enterprise under consideration.

5 Source: The new SME definition; user guide and model declaration (European Commission 92-894-7909-4).

22

4.1.2 Content of the quest ionnaire

The objective of the questionnaire was to obtain information on relevant indica-tors on the quality and quantity of jobs at enterprise level, and on the impact of (and the reaction to) the economic crisis that started in autumn 2008. Therefore, the questionnaire includes questions on the following topics (the final question-naire is included in Annex I): − general characteristics of the enterprise (sector, age, innovative behaviour); − general characteristics of the workforce (decomposition by age, educational

level and gender; employment of people with a physical or mental handicap); − indicators on quantity of jobs (number of employees, currently and twelve

months ago; working with employees from temporary work agencies, cur-rently and twelve months ago; employees laid off during the past twelve months; employees hired during the past twelve months; expected layoffs and hires for the next twelve months);

− indicators on quality of jobs (attention for training and other forms of life-long learning; main reasons for not providing training, if relevant; employees with part-time contracts; employees with fixed-term contracts; share of newly hired employees who were unemployed for at least one year);

− labour market position of the enterprise (factors making it easier or more dif-ficult to attract skilled employees);

− effects of the crisis (various negative and positive effects encountered during the past twelve months; layoffs due to the crisis; usage of publicly supported employment protection schemes).

The questionnaire was designed in close cooperation between EIM and the Euro-pean Commission. To allow for a comparison of the results with other studies, several questions were based on previous studies. These include the questions on innovative behaviour of enterprises, on the share of employees working full-time, on the workforce decomposition by age categories and on the workforce decomposition by educational level (based on the ISCED classification). These topics are now discussed briefly.

Innovation act ivi ties

The questionnaire contains three questions on innovation regarding product in-novation, process innovation and innovative activities during the past three years. These questions are aligned with the Community Innovation Survey (in particular; the text of the questions and additional explanations are based on the UK CIS 4 survey).

Employees working ful l -t ime

The distinction between fulltime and part-time employment differs between countries and within countries between sectors of industry. This makes it very difficult to establish a precise distinction in a multi-country, multi-sector survey. Following the EU Labour Force Survey, the Enterprise Survey therefore does not include an explicit threshold, but leaves it up to the interpretation of each re-spondent.

23

Decomposi t ion by age categories

One of the questions on the survey concerns the distribution of the workforce across three age categories. A common classification is to divide the working population into the following three age brackets1: − 15-24 years of age; − 25-49 years of age; − 50-64 years of age. These age brackets are also used in the Enterprise Survey 2010.

Educational levels according to ISCED classi f icat ion

Since its introduction in 1997 by UNESCO, the International Standard Classifica-tion of Education (ISCED) categorisation has been widely used as a framework for determining the educational level of a population. Amongst others, it is also used for the EU Labour Force Survey2. The ISCED classification covers six levels of educational (not counting the zero level, which represents no education what-soever): − ISCED 1: primary education or first stage of basic education; − ISCED 2: lower secondary education or second stage of basic education; − ISCED 3: (upper) secondary education; − ISCED 4: post-secondary non-tertiary education; − ISCED 5: first stage of tertiary education (not leading directly to an advanced

research qualification); − ISCED 6: second stage of tertiary education (leading to an advanced research

qualification). These six levels can be further classified into the following three main catego-ries3: − low: pre-primary, primary and lower secondary education (ISCED 0-2); − medium: upper secondary and post-secondary non-tertiary education (ISCED

3-4); − high: tertiary education (ISCED 5-6). These three categories form the framework for the Enterprise Survey 2010. Most entrepreneurs have no knowledge of this classification system. It is therefore not feasible to directly ask respondents which share of the workforce has an ISCED 0-2 education level (etc). It can be assumed, however, that respondents are fa-miliar with the education system in their country. For each country in the sam-ple, therefore, the national education system is linked to the ISCED classifica-tion. For most of the countries in our sample (including all EU Member States), this information is available through Eurybase, which is maintained by the Eury-

1 See, for example, the publications "Europe's demographic future" (DG Employment, Social Affairs

and Equal Opportunities, 2007) and "The flexibility of working time arrangements for women and men" (Eurostat Statistics in Focus, 2007).

2 See, amongst others, http://circa.europa.eu/irc/dsis/employment/info/data/eu_lfs/LFS_MAIN/Related_documents/ISCED_EN.htm.

3 See, for example, the background document on the indicator of youth education attainment level at http://circa.europa.eu/irc/dsis/employment/info/data/eu_lfs/LFS_MAIN/Related_documents/QP_Youth_Education_Level_Background%20document.htm.

24

dice Network1. For each of the countries in this database, the national education system is linked to the ISCED classification system. Eurybase does not contain information for the majority of the non-EU countries (Montenegro, Albania, Ser-bia, Croatia, Israel, and the Former Yugoslav Republic of Macedonia). Basic in-formation on the educational systems of these countries is available at euroedu-cation.net, and it is this information has been used. This information is used to determine country-specific lists of school types that represent the three main ISCED-based categories as defined above. For exam-ple, for Germany this results in the following examples: − low educational level: for example, grundschule, gesamtschule, realschule,

hauptschule; − medium educational level: for example, Gymnasiale oberstufe, fachober-

schule, berufsfachschule; − high educational level: for example, universität, fachhochschule, berufsa-

kademie. Likewise, for each country, country-specific examples of the various educational levels are included in the survey. The questionnaire included in the annex con-tains examples for the UK educational system. Finally, it should be noted that the questions pertain to the educational level of employees, not to their educational attainment. To the extent that life-long learning activities may increase the level of human capital of employees, it may also cause the educational level of employees to be higher than their educational attainment (which reflects the highest educational level obtained during initial education).

Test ing of the quest ionnaire

The final draft version was tested in the UK in early September 2010. After this pilot, several final changes were made to the questionnaire.

4.2 Stratification plan

This section describes the stratification plan used for this survey: the classifica-tions used; the objectives of the stratification plan; the sample plan mechanism applied to meet these objectives; the sources of the population data on which this mechanism has been applied; and the final sample plan.

4.2.1 Classi f icat ions

Countr ies

37 countries are distinguished: EU27 countries and 10 non-EU countries (Table 7). The total population amounts to 15 million employer enterprises2 in the busi-ness economy (defined in Section 4.2.2): 13 million (85%) in EU27, and 2 mil-

1 See http://eacea.ec.europa.eu/education/eurydice/eurybase_en.php.

2 In the remaining part of this chapter, when no confusion is possible, "employer enterprises" will be shortened to "enterprises".

25

lion in non-EU countries (over 50% of which in Turkey). Within EU27, 70% of the enterprises are located in the so-called 6 major economies (Poland, United King-dom, Germany, France, Spain, and Italy). Individual countries differ enormously with respect to the number of enterprises. On one side of the spectrum one finds Liechtenstein (1,600 enterprises) and Ice-land, Montenegro and Luxembourg (12,000-16,000 enterprises), on the other hand Spain (1.6 million) and Italy (2.4 million), implying a ratio between largest and smallest of over 1,400. This skewness of the distribution of the enterprises population is a key element in the sample design.

26

Table 7 Number of employer enterprises in the business economy, by countries (2008)

Country Number of employer enterprises

Poland 915,000

United Kingdom 1,212,000

Germany 1,442,000

France 1,516,000

Spain 1,575,000

Italy 2,356,000

6 major EU27 economies 9,014,000

Luxembourg 16,000

Malta 19,000

Estonia 28,000

Cyprus 30,000

Slovakia 43,000

Latvia 42,000

Ireland 59,000

Slovenia 60,000

Lithuania 76,000

Denmark 122,000

Finland 130,000

Bulgaria 160,000

Austria 184,000

Belgium 249,000

Romania 260,000

Sweden 286,000

Hungary 319,000

Netherlands 348,000

Greece 458,000

Portugal 470,000

Czech Republic 524,000

Other EU27 economies 3,882,000

EU27 12,896,000

Liechtenstein 1,600

Iceland 12,000

Montenegro 15,000

Albania 34,000

Serbia 50,000

Croatia 50,000

Norway 150,000

Israel 177,000

the Former Yugoslav Republic of Macedonia 386,000

Turkey 1,326,000

Non-EU 2,202,000

Total 15,099,000

Note: the business economy consists of NACE D, F-K, N and O (excl.91).

Source: EIM.

27

Sectors of industry

Within the business economy (NACE D, F -K, N, O excl. 91), seven main sectors are distinguished1. Within some of these main sectors (manufacturing, business services and personal services) a further distinction of elementary sectors is made: − Manufacturing, comprised of the following elementary sectors: metal industry;

food products, beverages, tobacco; pulp, paper and paper products; publish-ing and printing; textile and leather; electrical and optical equipment; wood and wood products and furniture; and other manufacturing (NACE Section D).

− Construction (NACE Section F). − Wholesale trade (NACE Division 51). − Retail trade, consisting of: sale, maintenance and repair of motor vehicles and

motorcycles; and retail sale of automotive fuel (NACE Divisions 50, 52). − Transport and communication (NACE Section I). − Business services, comprised of the following elementary sectors: financial in-

termediation; real estate activities; research and development, computer re-lated activities; other high-skilled business activities (accounting, consulting, market research, architectural and engineering activities, technical testing, advertising and recruitment services); and other low-skilled business activities (includes renting of machinery and equipment without operator and of per-sonal and household goods) (NACE Section J, K).

− Personal services, comprised of the following elementary sectors: hotels and restaurants; health and social work; and other community, social and per-sonal service activities (excluding activities of membership organisations) (NACE Sections H, N and O (excl. 91)).

The distribution of the stock of enterprises over sectors of industry is rather skew, the smallest elementary sector (electrical and optical equipment) having a population of 165,000, and the largest (retail trade) totalling 2.7 million enter-prises (Table 8).

1 NACE Rev. 1.1 is used for the sample plan and reporting.

28

Table 8 Number of employer enterprises in Europe-37, by sector (2008)

Sector

NACE rev. 1.1 Label

Number of

employer

enterprises

D Manufacturing 2,034,000

DJ Metal industry 359,000

DA Food products, beverages, tobacco 289,000

DE Pulp, paper and paper products; publishing and printing 201,000

DB, DC Textile and leather 255,000

DL Electrical and optical equipment 165,000

DD, DN Wood and wood products, furniture 370,000

DF, DG, DH, DI,

DK, DM

Other manufacturing 395,000

F Construction 2,013,000

51 Wholesale trade 1,526,000

50, 52 Retail trade 2,713,000

I Transport and communication 1,089,000

J, K Business services 2,892,000

J Financial intermediation etc. 199,000

70 Real estate activities 499,000

72, 73 Research and development, computer related activities 307,000

74.1 -74.5 Accounting, Consulting, market research, architectural

and engineering activities, technical testing, advertising

and recruitment services

1,414,000

74.6 -74.8, 71 Renting of machinery & equipment, other business ac-

tivities

474,000

H, N, O (excl. 91) Personal services 2,831,000

H Hotels and restaurants 1,122,000

N, O (excl. 91) Other service activities 1,708,000

D, F -K, N, O

(excl. 91)

Business economy 15,099,000

Source: EIM.

Size-classes

With respect to enterprise1 size, the following elementary classes are distin-guished: micro-enterprises, small and medium-sized enterprises, and large en-terprises (LSEs). Micro enterprises and small and medium-sized enterprises make up the SME-sector of the business economy. The distribution of the enter-prise population over size classes is skew, with 86% of the enterprises being mi-cro enterprises and 0.4% being LSEs (Table 9).

1 "Enterprises" is to be understood as employer enterprises. Statistically, they are defined as en-

terprises with more than one person employed.

29

Table 9 Number of employer enterprises in the business economy in Europe-37, by size

class (2008)

Size class Description Number of employer enterprises

Micro with staff 2-9 occupied persons 12,968,000

Small and medium-sized 10-249 occupied persons 2,076,000

LSEs >= 250 occupied persons 54,000

Total >= 2 occupied person 15,099,000

Note: the business economy consists of NACE D, F-K, N and O (excl.91).

Source: EIM.

4.2.2 Object ives of the strat i f ication plan

The survey should enable drawing inferences at a reasonably narrow confidence level for the following strata: 1. Micro enterprises, small and medium-sized enterprises, and LSEs at the level

of the business economy in each of the six major EU-countries. 2. SMEs in each elementary sector (and hence, also at the level of the main

sectors) in EU15 and EU12. 3. Micro enterprises, small and medium-sized enterprises and LSEs in each of

main sectors for EU15 and EU12. 4. SMEs in each individual country at the level of the business economy. 5. SMEs in each main sector for the aggregate of non-EU countries. 6. Micro enterprises, small and medium-sized enterprises and LSEs in each

elementary sector in EU27. There are two main issues to be considered: total sample size and the minimum number of observations required to draw reliable inferences in a stratum. There is a trade-off between these two: higher sample size reduces the variance of the weights to be used, while increasing the minimum number of observation re-quired to draw reliable inferences increases the variance of weights. The maxi-mum total sample size has been set at 7,500. Furthermore, it is assumed that "inferences at a reasonably narrow confidence level" can be drawn if the sample size equals 801. Furthermore, to prevent problems in obtaining sufficient contact details to exe-cute the survey, an overall restriction has been defined limiting sample size as less than 20% of the total enterprise population in each stratum. Additionally, it is known from experience that contact information is extremely difficult to obtain for financial intermediation (NACE J) and real estate activities (NACE 70). Be-cause of this, the maximum number of observations for these sectors of industry in EU27 is limited to 80 in each size-class (this is consistent with basic require-ment 6).

1 The estimated standard error of the point estimate for the estimation of π equals σ= (π⋅(1 -

π)/n)½, At given n, σ is maximum for π= 0.5. For n= 80 and π= 0.5, σ equals 0.055. With a prob-ability of 95%, the estimate would lie between 0.5 ± 1.96⋅0.055, or between 0.39 and 0.61.

30

4.2.3 Sample design mechanism

Suppose, for example, that (1) the smallest stratum for which inferences are to be made has a population of 1,500 (number of Liechtenstein SMEs, approxi-mately); (2) the total population amounts to 15 million (number of EU37 SMEs); and (3) reliable inferences can only be made with a minimum sample size of 100; then in case of a proportionate sample, the total sample size should equal 1,000,000 (100 times 15 million, divided by 1,500). This would make the survey prohibitively expensive1. Instead, one could limit the number of observations in the largest strata and conversely, have small strata over-represented in the overall design, and apply weighting to arrive at unbiased estimates at the aggre-gate level. This is known as disproportionate stratified sampling. However, weighting leads to an increase in the standard error of measurement for results at aggregate level2. It is intuitively clear that the more the distribution of observations resembles the population distribution over the strata, the smaller the increase in the standard error of estimate that results from weighting. This is equivalent to having weights wi as close to one as possible. In turn, this can be interpreted as the variance of the weights being lowest3. This can be achieved by an appropriate choice of the sample distribution. These arguments can be summarised as follows: − Stratification is necessary to meet all of the requirements for the sample

within reasonable costs. − Given the stratified nature of the sample, weighting is required to obtain un-

biased estimates at the aggregated level. − The sample design (i.e. the sample size by stratum) should be such that the

variance of weights is minimised. Once maximum sample size as well as other restrictions on the sample are given, the optimal sample plan can be calculated by solving a mathematical optimisa-tion model.

4.2.4 Sources of populat ion data

The main body of data is taken from the database developed in the framework of the Annual Report of the SME Performance Review 2009. It covers all 37 coun-tries in the sample. For most countries (including all EU Member States), the ref-erence year is 2008, but in some cases, the year closest to 2008 is chosen from the available data. The main database covers the non-financial business economy (NACE C -I, K) at the detail of NACE divisions. Additional estimates were prepared to arrive at the sectoral detail described in Section 4.2.1. This specifically refers to: − the inclusion of the Financial intermediation (NACE J); − the inclusion of Other service activities (NACE N, O excl. 91);

1 This is where the skewness issue (mentioned in Section 4.2.1) comes in: if the ratio between

smallest and largest stratum had been smaller, total sample size could have been smaller as well.

2 Under certain - but not unrealistic - conditions, it can be shown that samples with weights will have a higher variance than samples that do not need weights.

3 Obviously, were all wi's equal to one, their variance would equal zero.

31

− the further detail of Other business activities (NACE 74) into 74.1-74.5 and 74.6-74.8.

Also, non-employer enterprises had to be filtered out from the dataset.

Inclusion of f inancial intermediation and other serv ice act iv i ties

For EU-countries, Iceland, Liechtenstein, Norway and Turkey, this was based on data collected in the framework of the Observatory of European SMEs1. For other countries, extrapolations from EU-countries were made.

Further detai l in the other business act iv i ties sector (NACE 74)

It has been assumed that for each size class separately, the distribution of en-terprises over other business activities I and II is the same as in The Netherlands (the only country for which data on this subject has been identified).

Selecting employer enterprises

As the survey will be confined to employer enterprises, and the available data include non-employers as well, additional processing of the data is required. The number of employer enterprises has been estimated as follows: − A Pareto distribution2 has been fit through the distribution of enterprises over

micro, small, medium-sized and large enterprises for each sector of industry in each country.

− Using this distribution, the number of enterprises with 1, 2, …, 9 persons em-ployed has been calculated and subsequently fit against the known total num-ber of micro enterprises.

− Finally, the number of employers has been calculated as the number of enter-prises with at least two persons employed.

4.2.5 The f inal sample plan

The sample plan resulted in a stratified sample containing 1,998 different quo-tas. Table 10 shows the (aggregated) quota at the level of individual countries and size classes.

1 EIM Business and Policy Research: Observatory of European SMEs 2003, No. 7 (report submitted

to Enterprise Directorate-General of the European Commission). See http://ec.europa.eu/enterprise/policies/sme/files/analysis/doc/smes_observatory_2003_report7_en.pdf.

2 The Pareto distribution provides a simple model (only two parameters to be estimated) while still giving an acceptable description of the size-class distribution of enterprises in The Netherlands. See G. de Wit and J. de Kok: Enterprise size distributions in The Netherlands - their nature and their underlying employment distribution (in Dutch: Bedrijfsgrootteverdelingen in Nederland - Hoe zien zij er uit en hoe is de onderliggende werkgelegenheidsverdeling?); EIM, Research Re-port H200906, 2009; http://www.ondernemerschap.nl/index.cfm/12,html?nxt=ctm_publikatie&bestelnummer=H200906).

32

Table 10 Sample plan: number of enterprises by country and size class, business econ-

omy (NACE rev. 1.1 D, F -K, N, O, excl. 91)

Country

Micro

(2-9)

Small and medium-sized

(10-249)

Large

(250+) Total

Poland 319 87 175 581

United Kingdom 193 84 82 359

Germany 198 130 143 471

France 244 80 182 506

Spain 244 80 97 421

Italy 365 87 159 611

6 major EU27 economies 1,563 548 838 2,949

Luxembourg 19 61 19 99

Malta 23 127 7 157

Estonia 32 153 26 211

Cyprus 29 128 13 170

Slovakia 29 151 49 229

Latvia 25 143 39 207

Ireland 34 54 67 155

Slovenia 50 34 20 104

Lithuania 49 31 43 123

Denmark 49 44 99 192

Finland 57 35 93 185

Bulgaria 53 29 29 111

Austria 60 35 134 229

Belgium 78 30 29 137

Romania 60 25 40 125

Sweden 84 26 30 140

Hungary 74 19 32 125

Netherlands 85 33 25 143

Greece 108 23 23 154

Portugal 104 25 30 159

Czech Republic 134 23 91 248

Other EU27 economies 1,236 1,229 938 3,403

EU27 2,799 1,777 1,776 6,352

Liechtenstein 42 38 1 81

Iceland 24 56 14 94

Montenegro 38 42 14 94

Albania 39 41 7 87

Serbia 44 39 24 107

Croatia 62 18 15 95

Norway 57 23 18 98

Israel 60 21 17 98

the Former Yugoslav Republic

of Macedonia 75 23 11 109

Turkey 212 50 23 285

Non-EU 653 351 144 1,148

Total 3,452 2,128 1,920 7,500

Source: EIM.

33

4.3 Survey Interviews in the 37 countries concerned were made using questionnaires and native speakers in all relevant languages. The average length of the interviews varied by country and language; the French version was relatively long for in-stance. On average, though, the interviews took 20 minutes. After the pilot interviews were conducted and the final changes to the question-naire had been made, the actual fieldwork started at the end of September 2010. By December 31, 7,566 interviews had been completed. This exceeded the over-all quota of 7,500 interviews, but several individual quotas had not been met. During January and February 2011, 11 additional interviews were therefore con-ducted to ensure that enough observations were available in a few quota regard-ing "micro" and "small and medium-sized" enterprises. In the final sample most quotas of the sample plan were satisfactorily covered, in particular for micro and small and medium-sized enterprises (Table 11).

34

Table 11 Survey results: number of completed interviews by country and size class, busi-

ness economy (NACE rev. 1.1 D, F -K, N, O, excl. 91)

Country

Micro

(2-9)

Small and medium-

sized

(10-249)

Large

(250+) Total

Poland 318 94 170 582

United kingdom 188 72 80 340

Germany 135 113 108 356

France 246 84 181 511

Spain 217 106 109 432

Italy 372 89 160 621

6 major EU27 economies 1,476 558 808 2,842

Luxembourg 14 77 25 116

Malta 44 103 24 171

Estonia 46 153 21 220

Cyprus 52 99 22 173

Slovakia 42 147 41 230

Latvia 59 113 22 194

Ireland 62 77 63 202

Slovenia 50 40 16 106

Lithuania 40 60 27 127

Denmark 75 45 92 212

Finland 74 42 44 160

Bulgaria 56 40 31 127

Austria 59 41 77 177

Belgium 78 31 29 138

Romania 55 45 51 151

Sweden 78 38 31 147

Hungary 76 33 38 147

Netherlands 82 42 34 158

Greece 105 28 23 156

Portugal 91 39 29 159

Czech Republic 93 73 103 269

Other EU27 economies 1,331 1,366 843 3,540

EU27 2,807 1,924 1,651 6,382

Liechtenstein 52 28 0 80

Iceland 42 48 4 94

Montenegro 50 24 5 79

Albania 45 37 4 86

Serbia 58 49 36 143

Croatia 60 32 24 116

Norway 36 44 16 96

Israel 29 51 23 103

the Former Yugoslav Republic

of Macedonia 71 29 9 109

Turkey 154 109 26 289

Non-EU 597 451 147 1,195

Total 3,404 2,375 1,798 7,577

35

Source: EIM.

36

4.4 Preparation of the dataset

Plausibil ity check on effects of the cr is is

Upon completion of the fieldwork, two plausibility checks were performed on the data obtained: one regarding the answers to the positive and negative effects of the crisis, and one regarding the answers to the various questions regarding the quantity of jobs. The first check concerns the overall effects of the economic crisis. Two separate questions have been included in the survey concerning various positive and negative effects of the crisis (questions Q104 and Q106). These questions con-tain several subquestions, including a subquestion on whether or not the crisis had an overall negative or positive on the number of orders or total demand over the past two years. Since these questions refer to the net effect of the economic crisis on demand since the start of the crisis, respondents were expected to tick one of these options at the most, but not both. This was, however, not clear to all respondents: 6% of all respondents responded that the economic crisis had both an overall negative effect on the number of orders or total demand (ques-tion Q104_1) and an overall positive effect on the number of orders or total de-mand (question Q106_1). These inconsistent answers led to two adjustments to the data. First, for all re-spondents who answered "yes" to both questions Q104_1 and Q106_1, these an-swers have been recoded as "do not know". Second, for some of these respondents, the score on the variables concerning whether any positive or negative effects of the crisis had occurred had to be changed. Whether an enterprise encountered any negative (positive) effects from the crisis can be determined from the answers to all of the negative (positive) effects included in the questions: if none of the effects is reported, it is assumed that the enterprise was confronted directly with any negative (positive) effects of the crisis1. The two additional variables are named crisis_neg and crisis_pos. If the overall negative (positive) effect of the crisis on demand was the only nega-tive (positive) effect of the crisis, and if this overall negative (positive) effect was recoded as "do not know", then whether the crisis had any negative (posi-tive) effects on the enterprise is not known either. In these cases, the score on the variable crisis_neg (crisis_pos) is also recoded as "do not know". If, how-ever, the respondent also mentioned at least one other negative (positive) effect of the crisis, the conclusion remains that the crisis had a negative (positive) ef-fects on the enterprise. In these cases, the score on the variable crisis_neg (cri-sis_pos) does not need to be changed. For 99% of the respondents in the final sample, it is known whether or not the crisis had any negative (positive) effects on the enterprise; the answer category "do not know" occurs for 1% of the re-spondents.

1 This interpretation is valid because the last of the negative (positive) effects included in the

question is "any other negative (positive) effect". Thus, the list of possible negative (positive) effects of the crisis refers to all negative (positive) effects possible.

37

Plausibi l ity check on quest ions regarding quanti ty of labour

The second plausibility check concerns the internal consistency of the reported employment levels (now and twelve months ago) and on the reported numbers of employees laid off and hired. The total change in the employment level can be computed as the difference be-tween the current number of people employed and the number of people em-ployed twelve months ago. Changes in the employment level are often the result of laying off employees and/or hiring new employees. Nevertheless, the total change in employment level does not need to be equal to the sum of the number of employees laid off and hired during the past twelve months. An initial expla-nation for this difference is that the employment level may also change for rea-sons other than laying off or hiring employees. For example, employees may leave on their own account; they may leave because their temporary contracts are not renewed; or they may retire, become disabled or die. A second explana-tion is that the information obtained through the Enterprise Survey is an ap-proximation1. If the approximation is not exact, this will also result in differences between the sum of layoffs and hires on the one hand and the overall changes in employment levels on the other hand. Nevertheless, the information on the number of layoffs and hires allows for a few plausibility checks regarding the changes in the employment level during the past twelve months. First of all, typing errors may be detected by comparing the various employment variables. If the number of employees twelve months ago, plus the number of newly hired employees during the past twelve months, minus the number of employees laid off during the past twelve months is exactly 10 times higher (lower) than the current number of employees, it is very likely that the number of current employees written down contains a typing error. A similar argument can be made to detect likely typing errors in the number of employees twelve months ago. This procedure identified 5 likely typing errors (3 regarding current employment, 2 regarding employment twelve months ago). These typing errors have been corrected. Secondly, although individual enterprises may show large employment increases, in cases where the data indicate extreme employment growth, it is more likely that the data are wrong (because of a mistake by the respondent or by the inter-viewer) than that the employment growth of these enterprises was indeed so ex-treme. The difficulty is to determine when employment growth is so extreme. For this study, the following criteria have been applied. − For enterprises with no more than 100 employees, extreme growth is defined

as a Birch-corrected growth rate2 of 100 or more. A Birch-corrected growth rate of 100 occurs if an enterprise grows from 1 to 11 employees, from 5 to27

1 The respondents have to provide the answers immediately, without the opportunity to look them

up.

2 In the Birch-corrected growth rate, the standard growth rate is multiplied by a "correction fac-tor". The result is that a standard growth rate of 25% will be valued "higher" for larger enter-prises than for smaller enterprises. The correction factor used by Birch is the absolute difference of the number of workers from year 1 to year 2. The Birch-corrected growth rate has been used previously, amongst others in the European Observatories for SMEs. More details can be found in the annex to "European Observatory for SMEs; sixth report", published in 2000 by the European Commission.

38

employees, from 10 to 42 employees, from 20 to 65 employees or from100 to 200 employees.

− For larger enterprises, the Birch-correction is not required. Therefore, for en-terprises with 100 or more employees, extreme growth is defined as a growth rate of 1 or more (i.e. at least a doubling of the number of employees)1.

Besides employment growth, the data also include enterprises that report em-ployment contraction. The Birch-corrected growth rate cannot be used as a crite-rion to determine whether the employment reduction can be considered ex-treme2. Therefore, extreme employment reduction is defined in terms of the un-corrected employment growth rate: − For all enterprises, extreme employment contraction is defined as an employ-

ment growth rate of -0.9 or less (a reduction of 90% or more). According to these criteria, 30 enterprises in the sample showed an extreme em-ployment growth (extreme employment contraction did not occur). For these en-terprises, the increase in the employment levels was then compared to the re-ported numbers of employees laid off and hired during the past twelve months. If a considerable share of the increase in employment levels could be explained by the difference between the number of employees hired and laid off, it may be concluded that the extreme employment growth did indeed take place. If not, it may be concluded that the extreme employment growth is most likely erroneous. In this study, "a considerable share" is operationalised as 2/3. This means that if less than 33.3% of the increase in the employment level can be related to the net result of laying off and hiring employees, the information provided on the employment levels is considered too unreliable to use. In these cases, the obser-vations regarding the number of employees (both currently and twelve months ago) are recoded as missing. As a direct consequence, it is not possible to de-termine the proper weights for these enterprises3, which means that they are ef-fectively removed from the dataset. This is the case for 18 of the 30 enterprises with extreme employment growth.

Determining proper weights

Weights must be computed to correct for the disproportionate character of the sample. By using these weights, it becomes possible to present percentage dis-tributions that indeed represent the situation across the 37 European countries covered by the sample. The weights are computed using the actual distribution of the 15 million employer enterprises over the 1,998 quota that are distin-guished in the sample plan, based on the following three dimensions: − the three size classes; − the 18 sectors of industry; − the 37 countries.

1 For example, the Birch-corrected growth rate for an enterprise growing from 5,000 to 5,708

employees is also 100. Although this enterprise shows a considerable growth rate, it is not ex-treme.

2 The minimum Birch-corrected growth rate varies with firm size. For example, if an enterprise decreases from 100 to 10 employees, the Birch-corrected growth rate equals -81. When applied to a firm of 80 employees, this would result in a negative head count.

3 An exception can be made if the reported number of employees currently and twelve months ago are both in the same size class (for example, if employment increased from 2 to 9 employees, or from 260 to 5000). In these cases it is still possible to determine the proper weights, even if the information on the exact number of employees is not reliable. This is the case for 6 enterprises.

39

The final sample includes observations from 7,577 enterprises. However, be-cause reliable information on the size class is missing for 18 of these enterprises, weights have been computed for 7,559 enterprises. This can be considered as the size of the database used for the analyses in the main report.

Different iat ing between prevalence and magn itude

Some questions from the questionnaire ask for both prevalence and magnitude. For example, question Q152 (Does your enterprise currently have any vacancies? If so, how many?) combines two questions: first of all, whether or not the enter-prise currently has any vacancies (prevalence), and secondly (if the enterprise does have vacancies) how many (magnitude). The answers to these two ques-tions are recorded in a single variable (Q152), where "0" means "no" and a posi-tive number represents both "yes" to the question of whether the enterprise has any vacancies and the actual number of vacancies. It is often interesting to differentiate between the prevalence and the magnitude. To this end, an additional variable Q152_dich is computed; this is a dummy vari-able1 that registers whether or not an enterprise has any vacancies. Thus, two different basic tables can be determined: − regarding prevalence: a cross tabulation of Q152_dich by (e.g.) size class, in-

dicating the share of enterprises for each size class that had any vacancies; − regarding magnitude: average values of Q152 for (e.g.) different size classes,

indicating the average number of vacancies for each size class (for those en-terprises that reported having vacancies).

This dichotomisation has been done for 14 variables (Table 12).

1 "dich" stands for dichotomous.

40

Table 12 Dichotomisation of variables combining prevalence and magnitude

Survey Question Brief description

Newly defined

dummy variable

Q151 Do you currently employ employees from temporary work

agencies?

Q151_dich

Q160 Did you employ employees from temporary work agencies

twelve months ago?

Q160_dich

Q109_1 Have any employees been permanently laid off due to the

economic crisis?

Q109_1_dich

Q109_2 Have any employees been permanently laid off for other

reasons?

Q109_2_dich

Q111_1 Do you expect that employees will be permanently laid off

due to the economic crisis?

Q111_1_dich

Q111_2 Do you expect that employees will be permanently laid off

for other reasons?

Q111_2_dich

Q158_1 How many employees were hired to replace other employ-

ees who left the firm?

Q158_1_dich

Q158_2 How many employees were hired because of newly cre-

ated jobs?

Q158_2_dich

Q159 Had any of these newly hired employees been unemployed

for at least a year when they were recruited?

Q159_dich

Q117_1 How many employees will be hired to replace other em-

ployees who leave the firm?

Q117_1_dich

Q117_2 How many employees will be hired because of newly cre-

ated jobs?

Q117_2_dich

Q152 Does your enterprise currently have any vacancies? Q152_dich

Q153 Are some of these vacancies hard to fill? Q153_dich

Q161 Has your enterprise used any publicly supported employ-

ment protection scheme for any of your employees?

Q161_dich

Source: Enterprise Survey 2010, SMEs and EU Labour Market, EIM/GDCC (N=7559).

Combining absolute and relative answers

The questionnaire contains various questions about characteristics of the work-force of the enterprises. These questions generally ask for either an absolute number of employees or a share (or percentage). For example, question Q78 concerns the number of employees who are female. For enterprises with fewer than 50 employees, the question is formulated in terms of actual number of fe-male employees; for larger enterprises the question is formulated in terms of the share of female employees. These answers have to be made consistent before any analysis can be made. This has been done by calculating the correct per-centages for enterprises with fewer than 50 employees (by dividing the actual number of female employees by the total number of employees). The newly cre-ated variable has received a new name (female) and has been added to the dataset. This approach has been applied to 11 variables (Table 13).

41

Table 13 Defining auxiliary variables for variables containing absolute and relative an-

swers

Survey Question Brief description

Newly defined

relative variable

Q161

The share of employees who benefitted from publicly sup-

ported employment protection schemes Empl_protection

Q44 The share of employees with a fixed-term contract Fixed-term

Q48 The share of employees working on a full-time basis Fulltime

Q76_1 The share of employees with a low educational level Edu_low

Q76_2 The share of employees with a medium educational level Edu_medium

Q76_3 The share of employees with a high educational level Edu_high

Q77_1 The share of employees younger than 25 years of age Age_low

Q77_2 The share of employees between 25 and 50 years of age Age _medium

Q77_3 The share of employees aged 50 years or older Age _high

Q155 The share of people with a physical or mental handicap Disabled

Q78 The share of female employees Female

Source: Enterprise Survey 2010, SMEs and EU Labour Market, EIM/GDCC (N=7559).

Changing the order of the answer categories to quest ion Q121

Question Q121 was formulated as follows: "For each of the following aspects, can you indicate whether they make it easier or more difficult for your firm to attract skilled people compared to other firms, or whether there is no difference?" The answers have been recorded in a different order, to arrive at an ordinal vari-able with the following three answer categories: 1 = easier 2 = no difference 3 = more difficult.

A new indicator on innovativeness

The questionnaire includes three different questions on innovation (Q16, Q17 and Q107). The answers to these questions have been combined to generate a single indicator variable, indicating whether or not an enterprise can be consid-ered innovative (in a broad sense). Enterprises are considered innovative if they meet at least one of the following criteria: − the enterprise has introduced new or significantly improved goods or services

in the past three years; − the enterprise has introduced new or significantly improved processes for pro-

ducing or supplying goods or services in the past three years; − at least once a year the enterprise is engaged in activities to develop new

goods, services or production processes.

42

4.5 Analysis

Bivar iate analysis

Many of the figures and tables included in Chapters 6 and 8 to 12 of the publica-tion "Do SMEs create more and better jobs?" are based on the results of the En-terprise Survey 2010. For the majority, these figures and tables are based on bivariate analyses of the dataset. To this end, a large number of standard tables have been prepared, where the variables included in the dataset are tabulated against a limited number of control variables: − by size class; − by sector (main sectors as well as elementary sectors); − by country (individual countries as well as country groupings based on the in-

novation performance of countries1, competitiveness2 and income level3). In each case, two different tables have been prepared: a table containing an un-weighted number of enterprises (to determine whether enough observations are available to draw inferences at a reasonably narrow confidence level), and a ta-ble containing weighted percentage distributions (which provide an estimate of the population distribution). Depending on the measurement level of the variable of interest, the prepared table is a cross tabulation (in case of nominal or ordinal variables) or a table pre-senting averages per category (in case of scale variables)4. All in all, 100 (variables) x7 (control variables) x 2 (unweighted and weighted) = 1,400 standard tables have been prepared.

Multivariate analysis

In addition to the standard bivariate tables, a set of standard regression analy-ses has been performed. These regression analyses have been performed for al-most all questions included in the survey (the exceptions are the independence of the organisation, gender and position of the respondent, and sector and coun-try). Depending on the measurement level of the dependent variable, the regres-sion technique used is ordinary least squares (in case of a scale variable), a logit regression (in case of a nominal or dummy variable) or a multinomial logit re-gression (in case of an ordinal variable)5.

1 Countries are categorised into four groups based on the scores on the Innovation Union Score-

board 2010 (source: DG Enterprise and Industry).

2 Countries are categorised into two groups based on whether their score on the Global Competi-tiveness Index 2010 (source: World Economic Forum) is above or below the EU27 average.

3 Countries are categorised into four groups based on the Worldbank 2011 country classification by income. Because of the relatively high income levels in Europe, the lowest two income groups are empty.

4 The answer category "do not know/no answer" is included as a separate category in these tables.

5 The answer category "do not know/no answer" is excluded from these analyses.

43

The regression equations include the following independent variables: − Enterprise characteristics (firm size, firm age, sector, innovative behaviour). − Workforce characteristics (decomposition of the workforce by age, educational

level and gender; share of employees working fulltime; share of employees on a fixed-term contract).

− Country characteristics; these are incorporated in three different ways (each resulting in a separate regression analysis): − By including country dummies, differences between individual countries can

be identified that are not explained by country differences in the enterprise and workforce characteristics included in the regression model (available for all 37 countries).

− By including information on real GDP growth rates (for 2009 and 2010), the 2010 level of GDP per capital level (in purchasing power standards), and the national score on the Innovation Union Scoreboard 2010, it is possible to determine the extent to which country differences are related to the economic and innovation performance of countries. This information is available from Eurostat for 31 countries (these include all Member States plus Croatia, Iceland, Norway and Turkey).

− In addition to national statistics on GDP and innovation, several indicators from the country fact sheets from the Small Business Act have been in-cluded (these include the responsive administration index and the single market index). This information is available for 23 Member States (these indicators are not available for Cyprus, Malta, Bulgaria and Romania).

Results of these multivariate analyses were mentioned in the report whenever this was deemed relevant.

Cluster analysis: an enterprise typology on employment developments

Apart from the regression analyses, a different type of multivariate analysis has also been conducted. A cluster analysis has been performed in order to deter-mine an enterprise typology on employment developments. The cluster analysis was based on variables representing different aspects of employment developments such as laying off employees, recruiting new employ-ees, the employment growth rate and the usage of publicly supported employ-ment protection schemes. The results of the cluster analysis suggested an enter-prise typology of eight different enterprise types, each reflecting different com-binations of the actions of enterprises (hiring, firing, usage of support measures) and the outcomes of these actions (gross and net employment changes). Two of these types consist of enterprises where employment levels have re-mained the same, either because no employee flows occurred (the "stable" type), or because employees who left the firm were all replaced (the "replace-ments only" type). These enterprise types account for more than half of all en-terprises in the 37 countries included in the survey. Three other types ("fast growing", "hiring, no replacements" and "hiring with replacements") represent enterprises with (on average) positive employment developments. These three types account for 14% of the enterprise population. The remaining three enter-prise types ("supported", "setback" and "crisis") represent enterprises with (on average) negative employment developments and account for 33% of the enter-prise population.

44

The distribution of the enterprise types differs considerably across size classes. The main reason is that larger enterprises are less likely to report no labour flows during a year than smaller enterprises. For example, if more employees are employed, it comes increasingly likely that at least one employee leaves the firm on his or her own account or is laid off. As a result, the share of "stable" enter-prises decreases from approximately 50% for micro enterprises to 0% for large enterprises (Figure 1).

Figure 1 Enterprise typology of employment development, by size class, for EU37 (2010)

0%20%40%60%80%100%

Micro

Small and medium

Large

Total

Fast growing Hiring, no replacements Hiring, with replacements Replacements onlyStable Setback Supported Crisis

Source: Enterprise Survey 2010, SMEs and EU Labour Market, EIM/GDCC (N=7559); conducted

during the final quarter of 2010 (2010Q4).

The main purpose of deriving this typology was to examine the extent to which such a typology would be related to various indicators of job quality. This might help to identify which types of enterprises are more likely to provide jobs of higher quality (besides looking at size, country or sector). Various analyses have been performed to examine (cor)relations between this typology and available indicators for job quality, but these did not provide any additional insights. The enterprise typology is therefore not included in the main report.

45

ANNEX I Enterprise Survey 2010 Questionnaire

Question Applies to:

INTRODUCTION

TXT0 [Short Introduction - Mainly to use with GKs when they ask what the call concerns.] Good morning, my name is ________ from GDCC. I'm calling on behalf of the European Commission in Brussels and we'd like to dis-cuss employment trends and how they've been affected by the economic crisis. May I speak to the human resources manager, the general manager or the owner of your enterprise?

All

TXT1 [Full Introduction - To use primarily with suitable respondents] Good afternoon, this is________ from GDCC on behalf of the European Commission in Brussels. We're currently conducting a study throughout Europe to examine em-ployment trends and how these trends have been affected by the economic crisis. This is to enable the European Commission to better support companies in deliver-ing more and better jobs. Would you have some time to answer a few simple ques-tions? This interview will take approximately 20 minutes. Please note that the information you provide will be treated confidentially. It will not be used at an individual level nor handed over by name to the European Commis-sion or any other third party.

All

A: General characteristics of the enterprise and the respondent

Q1 Which of the following descriptions best characterises this enterprise? (INT.: read out) 1. An independent enterprise 2. An enterprise, where less than 25% of equity is owned by other enterprises 3. A subsidiary, or an enterprise where 25% or more of equity is owned by other

enterprises 4. Not an enterprise ⇒ END 1 999. (INT.: DO NOT READ) Do not know/no answer ⇒ END 1 INT: (clarification, do not read): Examples of organisations that are not enterprises are (semi-)government organisations, foundations, unions and charities.

All

Q2 What is your position within the enterprise? (INT.: read out) 1. Owner 2. HRM manager or Personnel manager 3. Director or General manager 4. Other member of the management team or group 5. Accountant 6. Family member of the owner(s) 7. Other, specify: … 999. (INT.: DO NOT READ) Do not know/no answer ⇒ END 1

All

Q3a INT.: note down gender of respondent without asking. 1. Male 2. Female 999. Do not know

All

Q4 How many establishments does your enterprise consist of? (INT: a reasonable estimate will do) 999. (INT.: DO NOT READ) Do not know/no answer

All

46

Q5 If Q4=1, the question should be formulated as follows: Approximately how many people are currently working for your enterprise? (Including yourself, fam-ily workers, other management and owners and regular employees, but ex-cluding temporary external workers). If (Q4>1 and Q4<999), the question should be formulated as follows: Approxi-mately how many people are currently working for your enterprise? (In-cluding yourself, family workers, other management and owners and regu-lar employees, also those in other establishments, but excluding temporary external workers). (INT: a reasonable estimate will do) (INT: All questions refer to NATIONWIDE employees, not outside the country we are calling!) 999999. (INT.: DO NOT READ) Do not know/no answer ⇒ END 1 If Q5=1 ⇒ END 1

All

Q6 And how many people were working for your enterprise twelve months ago? (INT: a reasonable estimate will do) 999999. (INT.: DO NOT READ) Do not know/no answer INT: (clarification, do not read): If the enterprise has been in existence for less than one year, the question refers to the start of the enterprise

All

Q7 What is the main activity of your enterprise? (INT.: ONLY ONE ANSWER ALLOWED, READ OUT IF NECESSARY) 1. Manufacturing ⇒ Q8 2. Construction: preparation, drilling, roads, buildings, installation,

plumbing, plastering, etc. ⇒ Q151 3. Retail trade, maintenance and repair of motor vehicles or personal &

household goods ⇒ Q151 4. Wholesale trade and commission trade except motor vehicles,

motorcycles ⇒ Q151 5. Transport, travel agencies, post & communications ⇒ Q151 6. Business services, incl. financial services and real estate ⇒ Q9 7 Hotels and restaurants ⇒ Q151 8 Other service activities ⇒ Q151 998. Other activities, i.e. agriculture, fishing, mining electricity, public

sector activities and defence, activities of membership organisations ⇒ END 1 999. (INT.: DO NOT READ) Do not know/no answer ⇒ END 1 INT: (clarification, do not read): Main in terms of turnover; largest share in enter-prise turnover INT: (clarification, do not read): Other service activities include health and social work, recreation, culture and sport activities and other service activities

All

Q8 What type of manufacturing? (INT.: ONLY ONE ANSWER ALLOWED, READ OUT IF NECESSARY) 1.1 Metal industry (basic metals, metal products except machinery and equip-

ment) 1.2 Food products, beverages, tobacco products 1.3 Pulp, paper and paper products, publishing, printing and recorded media 1.4 Textile and leather products 1.5 Electrical and optical equipment 1.6 Wood and wood products, furniture 1.7 Other manufacturing 999. (INT.: DO NOT READ) Do not know/no answer ⇒ END 1 ⇒ Q151

If Q7 =1

47

Q9 What type of business services? (INT.: ONLY ONE ANSWER ALLOWED, READ OUT IF NECESSARY) 6.1 Financial intermediation, insurance and pension funding, auxiliary activities 6.2 Real estate activities 6.3 Research and development, computer related activities 6.4 Accounting, consulting, market research, architectural and engineering activi-

ties, technical testing, advertising and recruitment services 6.5 Renting of machinery & equipment, other business activities 999. (INT.: DO NOT READ) Do not know/no answer ⇒ END 1

If Q7 =6

Q151 Do you currently employ employees from temporary work agencies? 0 no 1 (if yes) how many employees does this concern?____ (INT: fill in number of employees) 999998 (INT.: DO NOT READ) Do not know/no answer 999999 (INT.: DO NOT READ) Do not know the number of employees

All

Q160 And did you employ employees from temporary work agencies twelve months ago? 0 no 1 (if yes) how many employees did this concern?____

(INT: fill in number of employees) 999998 (INT.: DO NOT READ) Do not know/no answer 999999 (INT.: DO NOT READ) Do not know the number of employees INT: (clarification, do not read): If the enterprise has been in existence for less than one year, the question refers to the start of the enterprise

All

Q10 In which year did your enterprise start operating? 4 INT.: (Must be four digits, i.e. 1970 or 2001, reasonable estimate will do)

999. (INT.: DO NOT READ) Do not know/no answer

All

Q104 During the past two years, which of the following negative effects of the current economic crisis applied to your enterprise?

(INT.: read out; multiple answers allowed)

a an overall negative effect on the number of orders or total demand

b under-utilisation of the workforce

c increase in customers' payment terms

d bankruptcy or closure of major business partners

e shortage of working capital

f shortage of long term finance

g costs of finance (for example, interest rate)

h other

INT: (clarification, do not read): An overall negative effect on demand means that,

over the two-year period as a whole, the net effect of the economic crisis on de-

mand was negative; your enterprise may have gained some new clients or orders,

but the net effect should be negative.

INT: (clarification, do not read) A major business partner is another enterprise or

organisation with which the enterprise often collaborates in projects and/or activi-

ties.

INT: (clarification, do not read): If the enterprise has been in existence less than

two years, the question refers to the period since the start of the enterprise

All

48

Q106 During the past two years, which of the following positive effects of the cur-rent economic crisis applied to your enterprise? (INT.: read out; multiple answers allowed)

a an overall positive effect on the number of orders or total demand

b it is easier to hire skilled employees

c employees are willing to work more flexibly

d reduction in purchase prices

e lower wage costs

f it is easier to collaborate with other organisations

g other

INT: (clarification, do not read): An overall positive effect on demand means that,

over the two-year period as a whole, the net effect of the economic crisis on de-

mand was positive; your enterprise may have lost some new clients or orders, but

the net effect should be positive.

INT: (clarification, do not read): If the enterprise has been in existence less than

two years, the question refers to the period since the start of the enterprise

All

TXT3 INT: read out:

The following questions concern the innovation activities of your enterprise All

Q16 If (Q10<2007 or Q10=999), the question should be formulated as follows: Has your enterprise introduced new or significantly improved goods or services during the past three years?

If (Q10>2006), the question should be formulated as follows: Has your enterprise introduced new or significantly improved goods or services?

0 no

1 yes

9 (INT.: DO NOT READ) Do not know/no answer

INT: (clarification, do not read): By significant improvement, we refer to significant

improvements with respect to the capabilities of the good or service such as im-

proved user friendliness, components, software, or subsystems. The following im-

provements are excluded: a simple resale of new goods or services purchased from

other enterprises and changes of a solely aesthetic nature. The innovated good or

service (new or improved) must be new to your enterprise, but it does not need to

be new to your sector or market. It does not matter if the innovation was originally

developed by your enterprise or by other enterprises.

All

Q17 If (Q10<2007 or Q10=999), the question should be formulated as follows: Has your enterprise introduced any new or significantly improved processes for producing or supplying goods or services during the past three years?

If (Q10>2006), the question should be formulated as follows: Has your enterprise introduced any new or significantly improved processes for producing or supplying goods or services?

0 no

1 yes

9 (INT.: DO NOT READ) Do not know/no answer

All

49

INT: (clarification, do not read): This includes new or significant improvements to

the distribution method or support activities for your goods or services. Purely or-

ganisational innovations, however, are excluded. The innovated production process

(new or improved) must be new to your enterprise, but it does not need to be new

to your sector or market. It does not matter if the innovation was originally devel-

oped by your enterprise or by other enterprises.

Q107 How often is your enterprise engaged in activities to develop new goods, services, or production processes? Is this

(INT: read out; only one answer allowed)

1 never

2 less than once a year

3 less than once a month

4 less than once a week

5 every week

9 (INT.: DO NOT READ) Do not know/no answer

All

C: Quantity of jobs - employment changes

Q109_1 Q109_2

During the past twelve months, have any employees been permanently laid off? 0 no 1 yes (if yes) how many employees have been laid off due to the economic crisis ____

(INT: fill in number of employees) how many employees have been laid off for other reasons ____

(INT: fill in number of employees) (do not specify the "other reasons") INT: fill in 0 if answer is "no" INT: (clarification, do not read): Only permanent layoffs should be counted, not vol-untary retirements or temporary layoffs. Temporary layoffs occur when an enter-prise does not have enough work and therefore asks some of the workforce to stay at home. INT: (clarification, do not read): Other reasons for permanent layoffs include mis-conduct of employees at work (e.g. poor discipline, continually missing work), un-derperformance of employees, or a reorganisation of the enterprise not due to the economic crisis. INT: (clarification, do not read): If the enterprise has been in existence less than one year, the question refers to the start of the enterprise. INT: (clarification, do not read): This question refers to NATIONWIDE employees, not outside the country we are calling. 999998 (INT.: DO NOT READ) Do not know/no answer 999999 (INT.: DO NOT READ) Do not know the number of employees

All

Q111_1

Do you expect that employees will be permanently laid off in the next twelve months? 0 no 1 yes (if yes)

how many employees will be laid off due to the economic crisis? ____

(INT: fill in number of employees)

50

Q111_2 how many employees will be laid off for other reasons?____ (INT: fill in number of employees)

INT: fill in 0 if answer is "no" INT: (clarification, do not read): Only permanent layoffs should be counted, not vol-untary retirements or temporary layoffs. Temporary layoffs occur when an enter-prise does not have enough work and therefore asks some of the workforce to stay at home. INT: (clarification, do not read): Other reasons for permanent layoffs include mis-conduct of employees at work (e.g. poor discipline, continually missing work), un-derperformance of employees, or a reorganisation of the enterprise not due to the economic crisis. 999998 (INT.: DO NOT READ) Do not know/no answer

999999 (INT.: DO NOT READ) Do not know the number of employees

All

Q158_1 Q158_2

During the past twelve months, have any employees been hired? 0 no => Q117 1 yes (if yes) how many employees were hired to replace other employees who left the

firm?____ (INT: fill in number of employees) => Q159 how many employees were hired because of newly created jobs?____

(INT: fill in number of employees) => Q159 INT: (clarification, do not read): Enterprises may increase their labour force to gen-erate additional turnover. This may result in the creation of new jobs, for which em-ployees must be hired. These hires should be counted as "hired because of newly created jobs". Enterprises may also decide to hire individual employees if they be-lieve that these employees are valuable to their enterprise, even in the absence of an existing vacancy. These hires should also be counted as "hired because of newly created jobs". INT: (clarification, do not read): If the enterprise has been in existence less than one year, the question refers to the period since the start of the enterprise 999998 (INT.: DO NOT READ) Do not know/no answer => Q11

999999 (INT.: DO NOT READ) Do not know the number of employees => Q159

All

Q159 Had any of these newly hired employees been unemployed for at least a

year when they were recruited? 0 no 1 (if yes) how many employees did this entail?____ (INT: fill in number of employees) 999998 (INT.: DO NOT READ) Do not know/no answer

999999 (INT.: DO NOT READ) Do not know the number of employees

If Q158 =1 or Q158 =999999

Q117_1

Does your enterprise plan to hire new employees in the next twelve months? 0 no => Q161 1 yes (if yes)

how many employees will be hired to replace other employees who leave the

firm? (INT: fill in number of employees) => Q152

51

Q117_2 how many employees will be hired because of newly created jobs?____ (INT: fill in number of employees) => Q152

INT: (clarification, do not read): Enterprises may increase their labour force to gen-erate additional turnover. This may result in the creation of new jobs, for which em-ployees must be hired. These hires should be counted as "hired because of newly created jobs". 999998 (INT.: DO NOT READ) Do not know/no answer => Q161

999999 (INT.: DO NOT READ) Do not know the number of employees => Q152

All

Q152 Does your enterprise currently have any vacancies? 0 no => Q161 1 (if yes), how many vacancies does this entail?____ (INT: fill in number of vacancies) => Q153 999998 (INT.: DO NOT READ) Do not know/no answer => Q161 999999 (INT.: DO NOT READ) Do not know the number of vacancies => Q153

If Q117 =1 or Q117 =999999

Q153 Are some of these vacancies hard to fill? 0 no 1 (if yes) how many vacancies does this entail?____

(INT: fill in number of hard-to-fill vacancies) 999998 (INT.: DO NOT READ) Do not know/no answer 999999 (INT.: DO NOT READ) Do not know the number of hard-to-fill vacancies INT: (clarification, do not read): Hard-to-fill vacancies are all vacancies that you find hard to fill. There is no specific threshold associated with it.

If Q152 =1 or Q152 =999999

Q161 During the past two years, has your enterprise used any publicly supported employment protection scheme for any of your employees? 0 no 1 (if yes) how many employees/what share of employees did this entail?____

(INT: fill in number of employees (if Q5<50) or percentage of employees (if Q5>=50)

99999 (INT.: DO NOT READ) Do not know/no answer

INT: (clarification, do not read): Publicly supported employment protection schemes

are specific instruments aimed at helping employers retain their employees instead

of dismissing them, especially in times of economic distress. They may include, but

are not limited to, state allowances provided to employers in the form of temporary

payments or reductions of their compulsory social security contributions or meas-

ures allowing employers to use more flexible working-time arrangements.

All

52

D: Quality of job aspects

Q121 For each of the following aspects, can you indicate whether they make it easier or more difficult for your firm to attracted skilled people compared to other firms, or whether there is no difference? (answer categories: easier; more difficult; no difference; do not know/no answer) (INT.: read out; multiple answers allowed)

1 remuneration level

2 working-time arrangements

3 work-life balance

4 training opportunities

5 career opportunities

6 location of the enterprise

7 working climate in the enterprise.

8. exciting or challenging job profiles

9 other aspects

99 (INT.: DO NOT READ) Do not know/no answer

All

Q33 I will now mention several training and development activities. For each of these activities, can you indicate whether they have been used by none, some or many of your employees during the past twelve months? 1 firm-provided internal training courses

2 firm-provided external training courses

3 on-the-job training

4 mentoring programmes

5 job rotation

6 learning circles or quality circles

7 self-directed learning

8 attendance at conferences, workshops and lectures

9 study visits to other organisations

10 apprenticeships and traineeships

11 exchanges or secondments

12 other training and development activities

INT: (clarification, do not read) Self-directed learning is a training method where it

is not the teacher that initiates the learning, but the learner. The learner makes the

decisions about what training and development experiences will occur and how. The

learner selects and carries out his or her own learning goals, objectives, methods

and means to verify that the goals were met.

INT: (clarification, do not read) A learning circle or quality circle is a small group of

employees who meet regularly to study a relevant subject.

INT: (clarification, do not read) With exchanges or secondments, an employee gets

the opportunity to work for a certain period within another organisation.

INT: (clarification, do not read): If the enterprise has been in existence for less than

one year, the question refers to the period since the start of the enterprise.

If (answer to categories 1 and 2 is "none"), go to Q154;

Otherwise, go to TXT5

All

53

Q154 What are the main reasons that your enterprise did not provide training courses during the past twelve months? (INT.: read out; multiple answers allowed)

1 Employees have all the required skills

2 Training and development activities would not produce any benefits for the

enterprise

3 Financial costs of training

4 Lost working time while workers are being trained

5 Unable to cover work while workers are being trained

6 Lack of information about training opportunities

7 Can't find suitable external training and development

8 Lack of space or skills to provide internal training and development activities

9 Fear of trained workers leaving the enterprise

10 Lack of interest of employees in training and development activities

11 Other reasons

99 (INT.: DO NOT READ) Do not know/no answer INT: (clarification, do not read): If the enterprise has been in existence for less than one year, the question refers to the period since the start of the enterprise

answer to first two catego-ries of Q33 is "none"

E: Workforce characteristics

TXT5 INT: read out:

If Q4=1, the question should be formulated as follows: To conclude this survey, I will now ask you some questions regarding the workforce of your enter-prise. If (Q4>1 and Q4<999) To conclude this survey, I will now ask you some ques-tions regarding the workforce of your enterprise. This includes all of its es-tablishments

[INT: All questions refer to NATIONWIDE employees, not outside the country we are

calling!]

All

Instruc-tion

If Q5<50, all remaining questions will be in absolute numbers; If Q5>= 50, all remaining questions will be in shares/percentages, except for Q155

Q44 How many (what percentage) of the people working for your enterprise have a fixed-term contract? INT: Write down absolute number or percentage (0 to 100), depending on firm size 999. (INT.: DO NOT READ) Do not know/no answer

INT: (clarification, do not read): This question refers to employees who are hired by

the firm. Workers from temporary work agencies should not be included. (This also

applies to all of the remaining questions)

All

Q48 How many (what percentage) of the people working for your enterprise work on a full-time basis?

INT: write down absolute number or percentage (0 to 100), depending on firm size

999. (INT.: DO NOT READ) Do not know/no answer

All

Q76a How many (what share) of the people working for your enterprise have a lower educational level, for example, only primary school INT: write down absolute number or percentage (0 to 100), depending on firm size

99999. (INT.: DO NOT READ) Do not know/no answer

All

54

Q76b And how many have a medium educational level, for example, secondary school?

INT: Write down absolute number or percentage (0 to 100), depending on firm size

99999. (INT.: DO NOT READ) Do not know/no answer

All

Q76c And how many have a high educational level, for example, university? INT: Write down absolute number or percentage (0 to 100), depending on firm size

99999. (INT.: DO NOT READ) Do not know/no answer

All

Q77a How many (what share of) the people working for your enterprise are (is) younger than 25 years of age? INT: Write down absolute number or percentage (0 to 100), depending on firm size

99999. (INT.: DO NOT READ) Do not know/no answer

All

Q77b And how many are (what share is) between 25 and 50 years of age? INT: Write down absolute number or percentage (0 to 100), depending on firm size

99999. (INT.: DO NOT READ) Do not know/no answer

All

Q77c And how many are (what share is) 50 years of age or older? INT: Write down absolute number or percentage (0 to 100), depending on firm size

99999. (INT.: DO NOT READ) Do not know/no answer

All

Q155 Do you currently employ people with a physical or mental handicap? 0 no 1 (if yes) How many employees does this concern?____

(INT: Fill in number of employees) 999998 (INT.: DO NOT READ) Do not know/no answer

999999 (INT.: DO NOT READ) Do not know the number of employees

All

Q78 How many (what share) of the people working for your enterprise are fe-male?

INT: Write down absolute number or percentage (0 to 100), depending on firm size

99999. (INT.: DO NOT READ) Do not know/no answer

=>END2

All

END 1 Your company does not match the target group of this study. I would like to end this interview now. Thank you very much for giving me your time.

END 2 This was our last question. Thank you very much indeed for giving me so much of your time and providing useful information for policy making in Europe.